Diesel engine air system parameter optimization method, device, equipment and medium

Abstract

The invention discloses a diesel engine air system parameter optimization method, device and equipment and a medium. Comprising the steps that a leader module obtains engine state data and working condition data; judging whether a preset parameter updating condition is met or not based on the state data and the working condition data, and when the preset parameter updating condition is met, scheduling a follower module to execute parameter self-learning to obtain updated parameters; the leader module optimizes the air system controller based on the updated parameters to obtain an optimized controller; and performing control based on the optimized controller, and verifying a control effect. According to a diesel engine air system model predictive control architecture, a multi-parameter optimization agent algorithm is designed, so that in the full life cycle of an engine, a mechanism model, a data model or control parameters can be adaptively optimized according to the real-time performance degradation degree, redundant calculation is reduced, the learning efficiency is improved, and the system performance is improved. The method is suitable for complex and changeable actual roads or industrial scenes.

Description

Technical Field

[0001] This application relates to the field of engine control technology, and more specifically, to a method, apparatus, device, and medium for optimizing the parameters of a diesel engine air system. Background Technology

[0002] In the control algorithm of the engine air system, some parameters need to be tuned. The values ​​of these parameters directly affect the algorithm performance and control effect. Deviations in the values ​​can easily lead to overshoot, oscillation, and excessively long adjustment time in the control of boost pressure, and will also affect the tracking control of the target value to a certain extent.

[0003] Furthermore, as time goes on, engine deterioration occurs, such as air filter clogging, carbon buildup, and component aging. The original pressure and flow balance of the air system is disrupted, which in turn affects the combustion and after-treatment system of the diesel engine. Therefore, it is necessary to carry out self-learning of these parameters. Summary of the Invention

[0004] This application provides a method, apparatus, device, and medium for optimizing the air system parameters of a diesel engine, in order to at least solve the technical problem in the related art of the difficulty in adaptively learning multiple engine parameters.

[0005] According to one aspect of the embodiments of this application, a method for optimizing the parameters of a diesel engine air system is provided, comprising: The leader module acquires engine status data and operating condition data; Based on the status data and operating condition data, it is determined whether the preset parameter update conditions are met. If the preset parameter update conditions are met, the follower module is scheduled to perform parameter self-learning to obtain the updated parameters. Based on the updated parameters, the leader module optimizes the air system controller to obtain an optimized controller. Control is performed based on the optimized controller, and the control effect is verified.

[0006] In one implementation, the leader module acquires engine status data and operating condition data, including: The leader module collects engine status data and performs noise reduction processing on the status data; Based on a preset period, the denoised state data is analyzed for operating conditions. After the operating conditions are stable, the state data corresponding to the operating point is stored until multiple operating points of the engine are covered. The stored full-condition data is updated based on a preset time period.

[0007] In one implementation, based on the status data and operating condition data, it is determined whether a preset parameter update condition is met. If the preset parameter update condition is met, the follower module is scheduled to perform parameter self-learning to obtain the updated parameters, including: Based on the aforementioned state data, multiple preset deviations are monitored and accumulated in real time to obtain the cumulative deviation. When the cumulative deviation is greater than or equal to a preset threshold, the engine operating condition meets the preset stability condition, and the time since the last self-learning exceeds a preset period, it is determined that the parameter update condition is met. When the parameter update conditions are met, the first follower is scheduled to perform self-learning of the mechanistic model parameters, the second follower to perform self-learning of the data model parameters, and the third follower to perform self-learning of the model prediction control weight parameters.

[0008] In one implementation, when the parameter update condition is met, the first follower is scheduled to perform self-learning of mechanistic model parameters, the second follower to perform self-learning of data model parameters, and the third follower to perform self-learning of model prediction control weight parameters, including: When the cumulative deviation of the mechanism model is greater than or equal to the preset threshold, the engine operating condition meets the preset stability condition, and the time since the last self-learning exceeds the preset time period, the first follower is scheduled, and the mechanism model parameters are identified online using the recursive least squares method to obtain the mechanism model self-learning parameters. When the cumulative deviation of the data model is greater than or equal to the preset threshold, the engine operating condition meets the preset stability condition, and the time since the last self-learning exceeds the preset time period, the second follower is scheduled, and the radial basis function neural network is used to train the data model parameters to obtain the data model self-learning parameters. When the current cumulative control deviation is greater than or equal to the preset threshold, the engine operating condition meets the preset stability conditions, and the time since the last self-learning exceeds the preset time period, the third follower is scheduled, and the extreme value search algorithm is used to self-learn the model predictive control weight parameters to obtain the control weight parameters.

[0009] In one implementation, it further includes: Self-learning trigger conditions are configured based on mask calibration; or / and, It provides a manual trigger interface and enables self-learning of manual trigger parameters based on a dedicated diagnostic instrument.

[0010] In one implementation, the leader module optimizes the air system controller based on the updated parameters to obtain an optimized controller, including: The leader module obtains the updated mechanism model self-learning parameters, data model self-learning parameters, and control weight parameters; The updated mechanism model self-learning parameters, data model self-learning parameters, and control weight parameters are assigned to the air system mechanism model, data model, and model predictive control cost function, respectively, to obtain the optimized controller.

[0011] In one implementation, control is performed based on the optimized controller, and the control effect is verified, including: After the controller parameters are updated, each actuator will be controlled according to the updated parameters; Acquire actual operating status data of the engine air system during execution; The control effect is verified by comparing the actual operating status data with the system variable setpoints.

[0012] According to another aspect of the embodiments of this application, a diesel engine air system parameter optimization device is provided, comprising: The data processing module is used by the leader module to acquire engine status data and operating condition data; The parameter self-learning module is used to determine whether the preset parameter update conditions are met based on the status data and working condition data. When the preset parameter update conditions are met, the follower module is scheduled to perform parameter self-learning to obtain the updated parameters. The parameter update module is used by the leader module to optimize the air system controller based on the updated parameters, thereby obtaining an optimized controller. The status monitoring module is used to control the system based on the optimized controller and to verify the control effect.

[0013] According to another aspect of the embodiments of this application, an electronic device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to execute the above-described diesel engine air system parameter optimization method through the computer program.

[0014] According to another aspect of the embodiments of this application, a computer-readable storage medium is also provided, wherein a computer program is stored in the computer program, and the computer program is configured to execute the above-described diesel engine air system parameter optimization method when running.

[0015] The technical solutions provided in this application embodiment may include the following beneficial effects: The solution in this application embodiment, based on the model predictive control architecture of the diesel engine air system, designs a multi-parameter optimization intelligent agent algorithm. The multi-parameter optimization intelligent agent consists of two parts: a leader and a follower. The leader acquires engine status data and operating condition data, and determines whether the preset parameter update conditions are met based on the status data and operating condition data. When the parameter update conditions are met, the follower module is scheduled to perform parameter self-learning to obtain the updated parameters. The leader updates the learned core parameters to the air system controller, so that the engine air system can achieve the control of complex air systems and faster and more accurate target tracking effect under the control of the air system controller.

[0016] The intelligent scheduling mechanism based on the leader-follower collaborative architecture enables the system to adaptively optimize parameters according to the real-time performance degradation throughout the engine's entire life cycle, reducing redundant calculations, improving learning efficiency, and making it suitable for complex and ever-changing real-world road or industrial scenarios. Attached Figure Description

[0017] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a flowchart of a method for optimizing the parameters of a diesel engine air system according to an embodiment of this application; Figure 2 This is a schematic diagram of a model predictive control scheme for a diesel engine air system according to an embodiment of this application; Figure 3 This is a schematic diagram of a multi-parameter optimization agent algorithm according to an embodiment of this application; Figure 4 This is a schematic diagram of the division of an engine operating area according to an embodiment of this application; Figure 5 This is a schematic diagram of a diesel engine air system parameter optimization device according to an embodiment of this application; Figure 6 This is a schematic diagram of the structure of an optional electronic device according to an embodiment of this application. Detailed Implementation

[0018] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0019] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0020] Currently, existing solutions propose an ESO-MPC control structure, mainly consisting of three parts: a predictive model, an extended state observer, and an extreme value search. Model predictive control acquires the target quantity and real-time feedback state variables, continuously optimizing the solution for the optimal control law within each sampling period. The extended state observer, using control law information and the gas path state information of the controlled diesel engine, observes disturbances in real-time as disturbance terms in the predictive model and inputs them to the model predictive controller to compensate for model bias. The learner learns the optimal MPC (Model Predictive Control) weight coefficients over a long time scale based on the controller output and gas path state information.

[0021] The learner in this scheme only learns the MPC weight parameters in the model predictive controller. MPC control involves large-scale matrix operations, which puts significant strain on the processor. The more sophisticated the model, the greater the computational load, posing a challenge to real-time control. Furthermore, simplified models cannot guarantee accuracy, making it difficult to ensure effective control.

[0022] Based on this, embodiments of this application provide a method for optimizing the parameters of a diesel engine's air system. The method for optimizing the parameters of a diesel engine's air system according to embodiments of this application will be described in detail below with reference to the accompanying drawings. Figure 1 As shown, the method mainly includes the following steps: The S101 leader module acquires engine status data and operating condition data.

[0023] In this embodiment, the multi-parameter optimization agent consists of a leader and followers. During actual engine operation, the leader needs to flexibly adjust the followers' work according to specific circumstances to ensure the team's efficient and orderly operation. Therefore, the leader bears multiple responsibilities, including but not limited to data processing, task allocation, progress tracking, team management and coordination, decoupling control, and state updates.

[0024] First, leaders assume the role of data processing.

[0025] In one implementation, the leader module collects engine status data and performs noise reduction processing on the status data.

[0026] Specifically, leaders acquire raw data related to the engine's air system through system sensors or internal model calculations, such as TVA (Throttle Valve Actuator) opening, EGR (Exhaust Gas Recirculation) opening, VGT (Variable Geometry Turbocharger) opening, intake manifold pressure, exhaust manifold pressure, after-vortex pressure, intake manifold temperature, exhaust manifold temperature, ambient temperature, ambient pressure, engine speed, fuel injection quantity, torque, etc.

[0027] Furthermore, to eliminate noise and outliers introduced by factors such as sensor aging, environmental noise, and transmission interference, the leader uses filtering technology to denoise the raw data, retaining the main characteristics of the data and improving data reliability.

[0028] Furthermore, the denoised state data is analyzed based on a preset period. After the stable operating conditions are met, the state data corresponding to the operating point is stored until multiple operating points of the engine are covered.

[0029] Specifically, at regular intervals, leaders use the noise-reduced air system data in the current system for data analysis, including analysis of engine torque characteristics, fuel economy, emission characteristics, and power characteristics.

[0030] Furthermore, after the engine has met the operating conditions for a period of time, the stable operating conditions include: the rate of change of engine speed and the rate of change of fuel injection quantity within a specific range; the engine speed and fuel injection quantity within a specific range; the engine oil temperature within a specific range; the engine coolant temperature within a specific range; successful engine start exceeding a threshold; and no engine fault reporting. The leader begins storing the current system's data related to engine operating conditions and air system status variables. Storage of the same variable under the same operating condition is stopped when the data storage volume reaches a specific value, until the stored variable data covers... Figure 4 Data storage is stopped for the 81 operating conditions of the engine under low, medium, and high loads. The specific values ​​and ranges in this embodiment can be set according to actual circumstances.

[0031] Furthermore, the stored full-condition data is updated based on preset time periods.

[0032] To ensure the timeliness and representativeness of the data used for training the data model, the leader module employs a periodic data update mechanism. For example, every month, the stored full-condition data is updated. The system automatically clears all currently stored historical data and starts collecting and storing a new round of complete condition datasets from that moment. This proactive reset strategy effectively avoids the negative impact of outdated data on the model, ensuring that the model is always trained and compensated based on the latest and most accurate operating condition information that best reflects the current engine's actual state, thereby continuously providing a high-quality adaptive learning data source for the hybrid model.

[0033] S102 determines whether the preset parameter update conditions are met based on the status data and operating condition data. When the preset parameter update conditions are met, the follower module is scheduled to perform parameter self-learning to obtain the updated parameters.

[0034] In one implementation, firstly, multiple preset deviations are monitored and accumulated in real time based on state data to obtain the cumulative deviation; when the cumulative deviation is greater than or equal to a preset threshold, the engine operating condition meets the preset stability conditions, and the time since the last self-learning exceeds a preset time period, it is determined that the parameter update conditions are met.

[0035] The system leader monitors and accumulates deviations in the mechanistic model, data model, intake manifold pressure versus its setpoint, exhaust manifold pressure versus its setpoint, and EGR rate versus its setpoint in real time. If the accumulated deviation is below a preset threshold, it indicates that the current air system control law design is performing well, and the system does not need to enable parameter self-learning. If the accumulated deviation is greater than or equal to the threshold, it is considered that the current air system controller design has a large deviation, and the system needs to schedule the followers to perform self-learning based on the current engine operating conditions.

[0036] To facilitate the self-learning function of followers, the leader divides the engine's operating area into, for example: Figure 4 The 3x3 region shown in the star shape and Figure 4 The 9x9 area shown in the rectangle is used by the leader to determine the current engine location based on the filtered and valid data during engine operation. To prevent frequent switching between two adjacent areas, the leader sets a hysteresis loop when it detects a change in engine operating conditions. This means that the engine is only confirmed to have moved to another area after a certain period of time has elapsed since it left its current operating condition.

[0037] Furthermore, when the parameter update conditions are met, the first follower is scheduled to perform self-learning of the mechanistic model parameters, the second follower to perform self-learning of the data model parameters, and the third follower to perform self-learning of the model prediction control weight parameters.

[0038] Specifically, when the cumulative deviation of the mechanism model is greater than or equal to a preset threshold, the engine operating condition meets the preset stability conditions, and the time since the last self-learning exceeds a preset period, the first follower is scheduled, and the mechanism model parameters are identified online using the recursive least squares method to obtain the mechanism model self-learning parameters.

[0039] Mechanism models are mathematical models based on fundamental principles of physics and chemistry, capable of describing the inherent operating laws of a system. In engine control, mechanism models use physical laws such as mass conservation, energy conservation, and momentum conservation to establish mathematical equations for subsystems such as the intake system, combustion process, and emission generation, thereby achieving a physical description of the engine's operating process.

[0040] In one implementation, if the engine speed change rate and fuel injection quantity change rate are within a certain range, the engine oil temperature is within a specific range, the engine coolant temperature is within a specific range, the engine starts successfully beyond a threshold, and no engine fault is reported, then the engine is determined to be in a stable operating condition. Furthermore, if the cumulative deviation of the system's current operating point mechanism model exceeds a threshold and the time since the last self-learning in the current region exceeds the calibration time, then the parameter update conditions are determined to be met.

[0041] The learning flag position for the corresponding working condition of the mechanism model is 1. The leader schedules the first follower to start the self-learning of the mechanism model parameters. When the engine stays in the current area for more than a set value and the identification parameters learned by the RLS (Recursive Least Squares) algorithm stabilize at a certain value, the self-learning of the mechanism model for the current working condition area is completed.

[0042] The leader updates the MAP with the parameter input matrix B and perturbation matrix Q learned by RLS, sets the mechanistic model self-learning success flag to 1, and the first follower's learning ends. If the above conditions are not met, the mechanistic model self-learning success flag is set to 0, the self-learning function is turned off, the learning results for the current region are discarded, and the MAP for the corresponding region is not updated.

[0043] The pulse map (MAP) refers to a multidimensional data table stored in the ECU, used to store control parameters under various engine operating conditions, such as fuel injection quantity, ignition advance angle, and EGR rate. The self-learning algorithm updates the corresponding MAP table with the model parameters identified online, achieving adaptive correction of the control parameters.

[0044] Furthermore, when the cumulative deviation of the data model is greater than or equal to a preset threshold, the engine operating condition meets the preset stability conditions, and the time since the last self-learning exceeds a preset period, the second follower is scheduled, and the radial basis function neural network is used to train the data model parameters to obtain the data model self-learning parameters.

[0045] like Figure 4The illustrated operating conditions show that the 3×3 operating points of the mechanistic model are located at the center of the 9×9 operating points of the data model. Therefore, after the mechanistic model completes parameter self-learning for the corresponding operating condition region, the operating point regions of the data model surrounding the mechanistic model's operating points need to undergo parameter self-learning again.

[0046] Specifically, if the engine speed change rate and fuel injection quantity change rate are within a certain range, the engine oil temperature is within a specific range, the engine coolant temperature is within a specific range, and the engine starts successfully beyond a threshold; and no engine fault is reported, the engine is determined to be in a stable operating condition. Furthermore, if the cumulative deviation of the data model exceeds a certain threshold and the time since the last data model self-learning exceeds the calibration time, the parameter update conditions are determined to be met.

[0047] The leader schedules the second follower. Based on the RBF (Radial Basis Function) neural network, the data model parameters are trained using the operating condition data stored by the leader. After training, the parameters are updated in the corresponding MAP, and the data model self-learning success flag is set to 1, indicating that the current compensation value has been updated and the second follower's learning ends. If the above conditions are not met, the data model self-learning success flag is set to 0, and the current learning result is discarded.

[0048] Furthermore, when the current cumulative control deviation is greater than or equal to a preset threshold, the engine operating condition meets the preset stability conditions, and the time since the last self-learning exceeds a preset period, the third follower is scheduled, and the extreme value search algorithm is used to self-learn the model predictive control weight parameters to obtain the control weight parameters.

[0049] Since the mechanistic model parameters learned by the first follower are directly related to the data model parameters learned by the second follower and the control parameters learned by the third follower, if the mechanistic model parameters and data model parameters are identified, and the engine speed change rate and fuel injection quantity change rate are within a certain range, the engine oil temperature is within a specific range, the engine coolant temperature is within a specific range, the engine starts successfully exceeding a threshold, the engine reports no faults, and the current control deviation exceeds a certain threshold and the previous self-learning was not completed or the time since the last self-learning of control parameters exceeds a certain engine running time or mileage, then the parameter update conditions are determined to be met.

[0050] The leader directs the third follower to learn the weight parameters of MPC (Model Predictive Control) using an extreme value search algorithm. When the cost function of the extreme value search algorithm decreases to a specified threshold, the control model self-learning success flag is set to 1, the third follower completes the control parameter self-learning function, and updates the learned control increment weight matrix R to the corresponding MAP. If the above conditions are not met, the data model self-learning success flag is set to 0, and the current learning result is discarded.

[0051] In one implementation, self-learning trigger conditions are configured based on mask calibration; or / and, a manual trigger interface is provided to manually trigger parameter self-learning based on a dedicated diagnostic instrument.

[0052] In one implementation, the leader monitors various state variables in real time, such as the tracking error of intake pressure and EGR rate. Using preset thresholds and enabling conditions, it determines whether the primary source of current performance degradation is model mismatch (inaccurate mechanism or data model) or suboptimal control parameters. Based on the diagnostic results, the leader only enables the corresponding followers to perform optimization tasks. If the system response is slow or the overshoot is large, the third follower is triggered to learn control parameters; if the system performance is good, all followers remain dormant and do not perform any optimization.

[0053] Even when multiple objects need to be optimized, the leader coordinates in sequence, allowing followers to execute at different times and in different areas, thus avoiding computational conflicts and parameter coupling problems caused by simultaneous optimization.

[0054] In one implementation, the self-learning trigger condition can be selected via mask calibration before the engine rolls off the production line. During the after-sales phase, service stations can manually trigger the self-learning function using a diagnostic tool, flexibly addressing needs at different stages, such as parameter reset after fault repair. This reduces after-sales maintenance costs, supports remote or local parameter reset, avoids recall risks due to controller parameter drift, and improves the user experience.

[0055] The S103 leader module optimizes the air system controller based on the updated parameters, resulting in an optimized controller.

[0056] In one implementation, the leader module optimizes the air system controller based on the updated parameters to obtain an optimized controller. This includes: the leader module acquiring updated mechanistic model self-learning parameters, data model self-learning parameters, and control weight parameters; and assigning the updated mechanistic model self-learning parameters, data model self-learning parameters, and control weight parameters to the air system mechanistic model, data model, and model predictive control cost function, respectively, to obtain the optimized controller.

[0057] Specifically, the leader updates the mechanistic model parameters, data model parameters, and control parameters of the three followers in the MAP and assigns them to the air system mechanistic model, air system data model, and cost function in the air system control algorithm, respectively, to obtain the air system controller.

[0058] The controller uses the RLS algorithm to recursively identify model parameters and employs the GWO (Grey Wolf Optimizer) algorithm to optimize the forgetting factor offline, thereby achieving accurate identification of the core parameters of the mechanism model and reducing the difficulty and workload of calibrating the mechanism model in the control algorithm.

[0059] When the mechanistic model parameters obtained using the RLS algorithm with a forgetting factor are generalized to all operating conditions, the calculated intake manifold pressure, exhaust manifold pressure, and EGR rate values ​​deviate significantly from the actual values ​​under some conditions. An RBF neural network is used to compensate for the data in the dynamic models of intake manifold pressure, exhaust manifold pressure, and EGR rate, enabling data modeling and compensation for deviations in the unknown mechanistic components of the hybrid model. The ES algorithm is used to enable online real-time self-optimization of control parameters, allowing the control parameters to adapt to changes in operating conditions to achieve optimal control performance and improve controller robustness.

[0060] S104 uses the optimized controller for control and verifies the control effect.

[0061] In one implementation, control is performed based on an optimized controller, and the control effect is verified, including: after the controller parameters are updated, each actuator performs control according to the updated parameters; the actual operating state data of the engine air system during execution is acquired; and the control effect is verified by comparing the actual operating state data with the system variable setpoints.

[0062] Specifically, after the control parameters are updated online, each actuator operates according to the updated parameters. The EGR valve precisely controls the recirculated exhaust gas flow by adjusting its opening to achieve the EGR rate setpoint; the VGT adjusts the boost pressure by changing the turbine guide vane angle to track the exhaust manifold pressure setpoint; and the TVA adjusts the fresh air volume by controlling the throttle opening to achieve the intake manifold pressure setpoint. Under the updated control law, the actuators coordinate their actions, ensuring that the actual intake manifold pressure, exhaust manifold pressure, and EGR rate accurately track the setpoints.

[0063] During the setpoint tracking process, since the core parameters of the controller have been updated, the leader will obtain the updated actual state data of the engine air system using two methods to determine whether the current controller's control effect is ideal. One method is based on sensors, such as pressure sensors, temperature sensors, and flow meters, to collect actual state data. The other method is through model calculations, such as calculating manifold pressure model values, intake manifold temperature model values, fresh intake air volume model values, manifold temperature model values, EGR flow model values, and EGR rate model values ​​using the law of conservation of mass, fluid mechanics, and thermodynamics principles.

[0064] Furthermore, the control effect is verified by comparing actual operating state data with system variable setpoints. By comparing actual operating state data (such as intake manifold pressure, exhaust manifold pressure, and EGR rate) with system variable setpoints, the tracking performance of the controller can be quantitatively evaluated.

[0065] Understandably, when the deviation between the actual operating parameters (intake manifold pressure, exhaust manifold pressure, EGR rate) and the set values ​​exceeds a preset threshold, the system triggers a self-learning mechanism to form closed-loop control. This allows the system's actual state to accurately track the set values ​​again. This closed-loop self-learning mechanism of deviation detection, parameter update, and control optimization can continuously compensate for factors such as engine performance degradation and environmental changes, ensuring that the system maintains high control accuracy and robustness throughout its entire lifecycle.

[0066] To facilitate understanding of the methods in the embodiments of this application, the following description is provided in conjunction with the appendix. Figure 2 and attached Figure 3 Further description.

[0067] like Figure 2 As shown, this system adopts a closed-loop optimized control architecture. Taking the target values ​​of intake manifold pressure, exhaust pressure, and EGR rate as inputs, and based on a hybrid predictive model combining mechanism and data, the optimal actuator opening command is calculated through a dynamic decoupling module (including hybrid model feedback compensation and cost function rolling optimization) to drive the engine. The actual state value output by the engine is compared with the target value to form the main control loop. Simultaneously, the model disturbance observer estimates the total disturbance in real time and feeds it back to the dynamic decoupling module for feedforward compensation. Meanwhile, a multi-parameter optimization intelligent agent algorithm continuously learns online, feeding back updated mechanism, data, and control model parameters to the predictive model and controller, forming a parameter self-learning loop. The entire system achieves precise and adaptive decoupling control of a complex air system through a target tracking-disturbance compensation-parameter self-learning closed loop.

[0068] like Figure 3As shown, the multi-parameter optimization agent consists of a leader and followers. During engine operation, the leader, as the central decision-making unit, integrates four core functions: data processing, task allocation, status monitoring, and parameter updating, and uniformly manages the triggering, scheduling, and result integration of self-learning.

[0069] The leader records data through data processing functions, reflecting the actual movement of the air system. This data is then filtered and categorized to further analyze the system's operational status, and valid data is stored. Based on real-time data recordings, the leader calculates model deviations and analyzes follower enabling conditions, enabling sequence, and follower learning status. The tracking effect is evaluated based on set values ​​to determine whether and in what order the self-learning function should be activated. Finally, the leader judges the rationality of the learning results based on the learning effectiveness. If the data is unreasonable, the leader determines whether to relearn based on the actual strategy enabling conditions. If the follower's learning results meet expectations, the leader updates the learned core parameters in the air system controller, implementing parameter updates and monitoring the status.

[0070] The multi-parameter optimization intelligent agent follower mainly consists of three parts: follower ① mechanistic model parameter learning, follower ② data model parameter learning, and follower ③ control model parameter learning. Mechanism model parameter learning primarily uses the RLS algorithm to self-learn the parameters in the air system hybrid model, aiming to ensure the mechanistic model accurately reflects the dynamic process trends of the system. Data model parameter learning mainly uses the RBF neural network to self-learn the parameters in the air system hybrid model, aiming to ensure the state equation more accurately describes the system's equilibrium state. Control model parameter learning primarily uses the ES extreme value search algorithm to self-learn the control increment weight coefficients in the MPC cost function, aiming to ensure the system can still accurately calculate the engine's fresh air volume value even after problems such as engine aging occur, thus ensuring system emissions and fuel consumption. Furthermore, during the engine bench calibration phase, it can significantly reduce the workload of manual calibration; during vehicle operation, the ES algorithm continuously optimizes control parameters to adapt to long-term disturbances such as engine aging and changes in fuel quality.

[0071] The leader-follower collaborative architecture and optimization mechanism proposed in this solution can achieve the following significant beneficial effects: (1) Intelligent and efficient adaptive optimization: Through the unified scheduling of the leader, the system can intelligently select optimization objects (mechanism model, data model, control parameters) according to the performance degradation degree of real-time diagnosis throughout the engine's entire life cycle. This mechanism of "precise triggering, orderly execution, and partitioned updating" ensures the optimization effect while minimizing redundant calculations and significantly improving the overall efficiency of parameter learning, making it suitable for the real-time operating environment of vehicle ECU.

[0072] (2) Precise emission control and regulatory adaptability: Based on a hybrid dynamic modeling method of "mechanism model + data model", it combines the interpretability of physical laws with the data-driven compensation capability for complex nonlinearities and unknown mechanisms, enabling more accurate prediction of key emission-related parameters such as EGR rate and intake pressure. This directly translates into more precise feedforward and feedback control of NOx and PM emissions, providing a solid technical foundation for meeting stringent emission regulations and enhancing product compliance and market competitiveness.

[0073] (3) Reduced costs and improved robustness throughout the entire lifecycle: During the development and calibration phases, the collaborative framework of online self-learning (RLS, ES) and offline optimization (GWO) significantly reduces the workload of conducting numerous manual tests on the bench to calibrate complex model parameters, shortening the development cycle and reducing development costs. During vehicle operation and after-sales service, the system can continuously optimize online, adaptively compensating for long-term slow time-varying disturbances such as engine aging, component wear, and changes in fuel quality, extending the period during which the control system maintains optimal performance and improving long-term robustness. The dual-mode interface of mask calibration and manual triggering provides flexible and low-cost means for after-sales maintenance, fault recovery, and special operating condition handling, avoiding the risk of mass recalls that may be caused by parameter drift, significantly improving user experience and reducing maintenance costs throughout the entire lifecycle.

[0074] According to another aspect of the embodiments of this application, a diesel engine air system parameter optimization apparatus for implementing the above-described diesel engine air system parameter optimization method is also provided. For example... Figure 5 As shown, the device includes: Data processing module 501 is used by the leader module to acquire engine status data and operating condition data; The parameter self-learning module 502 is used to determine whether the preset parameter update conditions are met based on the status data and working condition data. When the preset parameter update conditions are met, the follower module is scheduled to perform parameter self-learning to obtain the updated parameters. The parameter update module 503 is used by the leader module to optimize the air system controller based on the updated parameters, thereby obtaining the optimized controller. The status monitoring module 504 is used to control based on the optimized controller and verify the control effect.

[0075] It should be noted that the diesel engine air system parameter optimization device provided in the above embodiments is only illustrated by the division of the above functional modules when executing the diesel engine air system parameter optimization method. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the equipment can be divided into different functional modules to complete all or part of the functions described above. In addition, the diesel engine air system parameter optimization device and the diesel engine air system parameter optimization method embodiments provided in the above embodiments belong to the same concept, and the implementation process is detailed in the method embodiments, which will not be repeated here.

[0076] According to another aspect of the present application, an electronic device corresponding to the diesel engine air system parameter optimization method provided in the foregoing embodiments is also provided to execute the diesel engine air system parameter optimization method described above.

[0077] Please refer to Figure 6 This illustrates a schematic diagram of an electronic device provided by some embodiments of this application. For example... Figure 6 As shown, the electronic device includes: a processor 600, a memory 601, a bus 602, and a communication interface 603. The processor 600, the communication interface 603, and the memory 601 are connected via the bus 602. The memory 601 stores a computer program that can run on the processor 600. When the processor 600 runs the computer program, it executes the diesel engine air system parameter optimization method provided in any of the foregoing embodiments of this application.

[0078] The memory 601 may include high-speed random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Communication between this system network element and at least one other network element is achieved through at least one communication interface 603 (which can be wired or wireless), such as the Internet, wide area network, local area network, or metropolitan area network.

[0079] Bus 602 can be an ISA bus, PCI bus, or EISA bus, etc. Buses can be divided into address buses, data buses, control buses, etc. Memory 601 is used to store programs. After receiving execution instructions, processor 600 executes the programs. The diesel engine air system parameter optimization method disclosed in any of the aforementioned embodiments of this application can be applied to processor 600, or implemented by processor 600.

[0080] The processor 600 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of the processor 600 or by instructions in software form. The processor 600 may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), an off-the-shelf programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules may reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in memory 601. Processor 600 reads the information in memory 601 and, in conjunction with its hardware, completes the steps of the above method.

[0081] The electronic device provided in this application embodiment and the diesel engine air system parameter optimization method provided in this application embodiment are based on the same inventive concept and have the same beneficial effects as the methods they adopt, operate or implement.

[0082] According to another aspect of the present application, a computer-readable storage medium corresponding to the diesel engine air system parameter optimization method provided in the foregoing embodiments is also provided, wherein a computer program (i.e., a program product) is stored thereon, and when the computer program is run by a processor, it executes the diesel engine air system parameter optimization method provided in any of the foregoing embodiments.

[0083] It should be noted that examples of computer-readable storage media may also include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other optical and magnetic storage media, which will not be elaborated here.

[0084] The computer-readable storage medium provided in the above embodiments of this application and the diesel engine air system parameter optimization method provided in the embodiments of this application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the application programs stored therein.

[0085] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0086] The above embodiments merely illustrate several implementation methods of the present invention, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this patent should be determined by the appended claims.

Claims

1. A method for optimizing the parameters of a diesel engine's air system, characterized in that, include: The leader module acquires engine status data and operating condition data; Based on the status data and operating condition data, it is determined whether the preset parameter update conditions are met. If the preset parameter update conditions are met, the follower module is scheduled to perform parameter self-learning to obtain the updated parameters. Based on the updated parameters, the leader module optimizes the air system controller to obtain an optimized controller. Control is performed based on the optimized controller, and the control effect is verified.

2. The method according to claim 1, characterized in that, The leader module acquires engine status data and operating condition data, including: The leader module collects engine status data and performs noise reduction processing on the status data; Based on a preset period, the denoised state data is analyzed for operating conditions. After the operating conditions are stable, the state data corresponding to the operating point is stored until multiple operating points of the engine are covered. Full-condition data is updated and stored based on preset time periods.

3. The method according to claim 1, characterized in that, Based on the status data and operating condition data, it is determined whether the preset parameter update conditions are met. If the preset parameter update conditions are met, the follower module is scheduled to perform parameter self-learning to obtain the updated parameters, including: Based on the aforementioned state data, multiple preset deviations are monitored and accumulated in real time to obtain the cumulative deviation. When the cumulative deviation is greater than or equal to a preset threshold, the engine operating condition meets the preset stability condition, and the time since the last self-learning exceeds a preset period, it is determined that the parameter update condition is met. When the parameter update conditions are met, the first follower is scheduled to perform self-learning of mechanistic model parameters, the second follower to perform self-learning of data model parameters, and the third follower to perform self-learning of model prediction control weight parameters.

4. The method according to claim 3, characterized in that, When the parameter update conditions are met, the first follower is scheduled to perform self-learning of mechanistic model parameters, the second follower to perform self-learning of data model parameters, and the third follower to perform self-learning of model prediction control weight parameters, including: When the cumulative deviation of the mechanism model is greater than or equal to the preset threshold, the engine operating condition meets the preset stability condition, and the time since the last self-learning exceeds the preset time period, the first follower is scheduled, and the mechanism model parameters are identified online using the recursive least squares method to obtain the mechanism model self-learning parameters. When the cumulative deviation of the data model is greater than or equal to the preset threshold, the engine operating condition meets the preset stability condition, and the time since the last self-learning exceeds the preset time period, the second follower is scheduled, and the radial basis function neural network is used to train the data model parameters to obtain the data model self-learning parameters. When the current cumulative control deviation is greater than or equal to the preset threshold, the engine operating condition meets the preset stability conditions, and the time since the last self-learning exceeds the preset time period, the third follower is scheduled, and the extreme value search algorithm is used to self-learn the model predictive control weight parameters to obtain the control weight parameters.

5. The method according to claim 1, characterized in that, Also includes: Self-learning trigger conditions are configured based on mask calibration; or / and, It provides a manual trigger interface and enables self-learning of manual trigger parameters based on a dedicated diagnostic instrument.

6. The method according to claim 1, characterized in that, Based on the updated parameters, the leader module optimizes the air system controller to obtain an optimized controller, including: The leader module obtains the updated mechanism model self-learning parameters, data model self-learning parameters, and control weight parameters; The updated mechanism model self-learning parameters, data model self-learning parameters, and control weight parameters are assigned to the air system mechanism model, data model, and model predictive control cost function, respectively, to obtain the optimized controller.

7. The method according to claim 1, characterized in that, Control is performed based on the optimized controller, and the control effect is verified, including: After the controller parameters are updated, each actuator will be controlled according to the updated parameters; Acquire actual operating status data of the engine air system during execution; The control effect is verified by comparing the actual operating status data with the system variable setpoints.

8. A device for optimizing air system parameters of a diesel engine, characterized in that, include: The data processing module is used by the leader module to acquire engine status data and operating condition data; The parameter self-learning module is used to determine whether the preset parameter update conditions are met based on the status data and working condition data. When the preset parameter update conditions are met, the follower module is scheduled to perform parameter self-learning to obtain the updated parameters. The parameter update module is used by the leader module to optimize the air system controller based on the updated parameters, thereby obtaining an optimized controller. The status monitoring module is used to control the system based on the optimized controller and to verify the control effect.

9. An electronic device, characterized in that, It includes a processor and a memory storing program instructions, the processor being configured to perform the diesel engine air system parameter optimization method as described in any one of claims 1 to 7 when executing the program instructions.

10. A computer-readable medium, characterized in that, It stores computer-readable instructions that are executed by a processor to implement a diesel engine air system parameter optimization method as described in any one of claims 1 to 7.

Images